The time-averaged shear-wave velocity in the upper 30 m of a site (VS30) is commonly used for ground-motion prediction. When measured velocities are unavailable, VS30 is estimated from proxy-based relationships developed for application on global or local scales. We describe the development of a local relationship for Greece, which begins with compilation of a profile database (PDB) from published sources and engineering reports. The PDB contains 314 sites; 238 have profile depths ≥ 30 m and 59 are within 100 m of accelerographs. We find existing relations for extrapolating a time-averaged velocity for depths less than 30 m to VS30 to overpredict VS30. We present equations for these extrapolations. We then compile proxies for PDB sites, including terrain type, surface geology, and surface gradients at 30 and 3 arcsec resolution (from radar-derived digital elevation models [DEMs]). When checked against ground survey data, we find ground elevations from 3 arcsec DEMs to be more accurate relative to survey data than alternative 30, 9, and 1 arcsec DEMs. Drawing upon expert opinion, we develop geologic categories based on age, gradation, and depositional environment and assign such categories to PDB sites. We find an existing 30 arcsec gradient-based global model to be biased relative to local VS30 data for gradients >∼0:05 m=m. Bias relative to a California model is also found for four of the eight well-populated geomorphic categories, and new (local) values are provided.We find statistically significant effects of the 3 arcsec gradient on VS30 for Quaternary and Tertiary materials but no gradient effect for those from the Mesozoic. Among Quaternary sediments, Holocene, mapped Quaternary (age unspecified), and mixed/fine-gradation materials exhibit consistent VS30-gradient trends, whereas Pleistocene and coarse-gradation sediments have faster velocities. For the study region, we recommend use of the modified terrain- and geology- based methods in combination for proxy-based VS30 estimation.